Are Cisco and Lake Whitefish Competitors? An Analysis of Historical Fisheries in Michigan Waters of the Upper Laurentian Great Lakes

Historically, Cisco Coregonus artedi and Lake Whitefish Coregonus clupeaformis were abundant throughout the Laurentian Great Lakes (hereafter Great Lakes; Koelz 1929), but overharvest, habitat degradation, and interactions with exotic species caused most populations to collapse by the mid-1900s (Fleischer 1992; Madenjian et al. 2008; Gorman 2012). In lakes Superior, Michigan, and Huron (hereafter upper Great Lakes; Figure 1), harvests declined 93–100% from historical peaks between the late 1800s and early 1900s to all-time lows during the 1950s and 1960s (Figure 2; Baldwin et al. 2018). Strict commercial fishery regulations (e.g., Brege and Kevern 1978) and a few strong year-classes (Bronte et al. 2003) enabled Cisco to partially recover in Lake Superior (Rook et al. 2021) and harvests subsequently increased after the early 1980s (see Figure 2). Despite environmental recovery in historically important Cisco spawning and rearing areas, such as Green and Saginaw bays (see Figure 1; Madenjian et al. 2008, 2011), remnant stocks in lakes Michigan and Huron showed limited signs of recovery and permitted harvests remained low or not allowed (Ebener 2013; Claramunt et al. 2019). By contrast, Lake Whitefish recovered throughout all three lakes (Fleischer 1992; Ebener et al. 2008b; Madenjian et al. 2008) and harvests subsequently increased to record levels after the 1980s, although there have been declines in recent years, especially in lakes Michigan and Huron (see Figure 2). Effective Sea Lamprey Petromyzon marinus control (Madenjian et al. 2008) and a series of cold winters with warm spring temperatures that created ideal conditions for recruitment (e.g., Freeberg et al. 1990) likely drove the recovery of Lake Whitefish in all three lakes. Cisco and Lake Whitefish populations increased simultaneously in Lake Superior during 1978–1998 (Bronte et al. 2003), but differential responses to improved environmental and ecological conditions in lakes Michigan and Huron during the same period have led to questions about potential negative interactions between these species. For example, when discussing restoration options for both species in Lake Erie, Oldenburg et al. (2007) suggested stocking Cisco and Lake Whitefish at different locations to reduce potential competition. Similarly, the Lake Huron Technical Committee (LHTC) of the Great Lakes Fishery Commission discussed competition between these species when considering restoration options for Cisco, but concluded the likelihood of competition was “not great” (LHTC 2007). Commercial fishermen also routinely voice concerns about potential negative impacts of Cisco reintroduction on commercially valuable Lake Whitefish at stakeholder meetings around Lake Michigan (C. Bronte, U.S. Fish and Wildlife Service [USFWS], personal observation).

The perception that Cisco and Lake Whitefish may interact to the detriment of one another in the upper Great Lakes has its basis in trends in annual harvest since the 1970s (see Figure 2) and results of mesocosm experiments that suggest competition for limited zooplankton food resources could be important during the larval stage (e.g., Todd and Davis 1995). However, the extirpation of Cisco throughout most of their former range in lakes Michigan and Huron (see Eshenroder et al. 2016) likely occurred at least 10–20 y before the recovery of Lake Whitefish (see Figure 2), which suggests no causal relationship between increased Lake Whitefish abundance and Cisco declines. Mesocosm experiments also nearly always result in greater competitive effects than experiments with free-ranging animals and may not represent “natural” conditions (Gurevitch et al. 1992). Coexistence of Cisco and Lake Whitefish has occurred for thousands of years (Bailey and Smith 1981), and before the mid-1900s, spawning and nursery areas throughout the upper Great Lakes supported abundant stocks of both species (Koelz 1929; Goodyear et al. 1982). Lake Whitefish evolved an inferior mouth and are benthic, as opposed to all other Coregonus forms in the Great Lakes that evolved terminal or superior mouths and are pelagic or bentho-pelagic (Koelz 1929; Becker 1983; Eshenroder et al. 2016), suggesting some adaptation to reduce competition among closely related forms within the genus (e.g., Smith and Todd 1984). Segregation of adults of each species generally occurs in space, so they are unlikely to be strong competitors (pelagic Cisco vs. benthic Lake Whitefish; Becker 1983). Similarly, although larvae of both species are pelagic (Becker 1983), most field studies suggest that Cisco larvae are found over deeper offshore waters, whereas Lake Whitefish larvae are generally found in shallower nearshore habitats (e.g., McKenna et al. 2020). In areas where Cisco spawn in western Lake Superior, estimates of consumption of available Cisco eggs by adult Lake Whitefish were substantial (41%; Stockwell et al. 2014), but the implications of this finding for recruitment of adult Cisco are unclear. Evidence from inland lakes suggests a relationship between recruitment bottlenecks for Lake Whitefish and Cisco predation on Lake Whitefish larvae (e.g., Carl and McGuiness 2006), but evidence from the Great Lakes includes only general accounts of Cisco piscivory on other species (e.g., Link et al. 1995).

There has been a decline in total phosphorus levels in the upper Great Lakes since the early 1970s (Dove and Chapra 2015), thereby reducing overall productivity and zooplankton abundance and changing zooplankton species and size composition, especially since the mid-2000s (e.g., Barbiero et al. 2012). These changes likely reduced production of coregonines, as previously observed in northern European lakes (Eckmann et al. 2007). However, current trophic conditions in lakes Michigan and Huron are similar to those of ultraoligotrophic Lake Superior (total phosphorus levels ≤ 4 μg/L; Dove and Chapra 2015), which continues to support commercial harvests of both Cisco and Lake Whitefish (Baldwin et al. 2018). Diatom-inferred total phosphorus levels suggest that current trophic conditions in Lake Superior may represent a more “natural” presettlement state (Shaw-Chraibi et al. 2014) wherein both Cisco and Lake Whitefish were abundant (e.g., Thwaites 1899). Colonization by nonnative dreissenid mussels Dreissena spp. has also been limited in Lake Superior and native food webs have remained largely intact (Ives et al. 2019). By contrast, large-scale ecosystem changes related to the proliferation of dreissenids may have permanently altered energy pathways and habitat preferences in lakes Michigan and Huron (Nalepa et al. 2005; Ebener et al. 2008b; Ives et al. 2019). As a result, competitive interactions between Cisco and Lake Whitefish could intensify under more recent environmental and ecological conditions in both lakes, especially during the larval stage (e.g., Todd and Davis 1995).

To provide context for fishery managers, we tested for positive and negative correlations between historical (1929–1970) Cisco and Lake Whitefish commercial gill net catch per effort (CPE; kilograms per kilometer of net; see Data S1, Supplemental Material) at a variety of spatial scales in Michigan waters of the upper Great Lakes (see Figure 1). We used the 42-y historical period to provide a first look at potential interactions between Cisco and Lake Whitefish following ongoing Cisco reintroduction efforts in the upper Great Lakes (see Bronte et al. 2017). We acknowledge that large-scale ecosystem changes since the mid-2000s may limit potential coregonine production compared with conditions during 1929–1970 (e.g., Eckmann et al. 2007). However, contemporary data were unavailable to test for interactions between Cisco and Lake Whitefish because of widespread Cisco extirpations throughout most of the Great Lakes after the mid-1900s (see Eshenroder et al. 2016). Similarly, fishery-independent data were unavailable to test for interactions between these species because annual fishery-independent assessments were not implemented until the late-1970s (e.g., Gorman 2019). In the absence of fishery-independent data, previous studies commonly used standardized fishery-dependent data to estimate abundances and provide management context for many freshwater and marine fish species (e.g., Hilborn et al. 2020; Palomares et al. 2020). Our analyses required the assumption of a constant one-to-one relationship between commercial gill net CPE and underlying abundance for both Cisco and Lake Whitefish. Although commonplace, this assumption was likely incorrect because of the presence of hyperstability (see Harley et al. 2001) and spatial differences in catchability and the way commercial fisheries operate (see Walters 2003). However, because the relative difference between Cisco and Lake Whitefish CPE was most important for identifying potential interactions between these species, we assumed that a meta-analytic approach based on a wide range of contrasting CPE estimates for each species would allow significant correlations to transcend any deficiencies in the data (e.g., Skud 1982). Similarly, we assumed that pairing Cisco and Lake Whitefish CPE at the smallest spatial scale possible, individual commercial fishery management units (see Figure 1), would account for any spatial differences in the data (e.g., Walters 2003).

Historical gill net fisheries for Cisco and Lake Whitefish operated separately from each other in both time and space. Commercial fishers targeted Cisco with 50.8–76.2-mm (2.0–3.0-in.) stretch-measure gill nets (Van Oosten 1929; Smith 1956; Dryer and Beil 1964), whereas commercial fishers targeted Lake Whitefish with mostly 114.3-mm (4.5-in.) stretch-measure gill nets (Jensen 1976). We used annual totals of Cisco and Lake Whitefish catch (kilograms) and effort (kilometers of net) data from 1929 to 1970 (Jensen and Buettner 1976) to estimate CPE for 20 commercial fishery management units (Smith et al. 1961) in state of Michigan waters of lakes Superior, Michigan, and Huron (see Figure 1). CPE estimates for both species were not corrected for fishing time (number of nights) because the required data were unavailable and annual changes in estimates of CPE and CPE corrected for fishing time were previously shown to differ insignificantly from each other (Hile 1962). Linen and cotton gill nets were used early in the time series (Cisco and Lake Whitefish), followed by nylon-multifilament (Cisco and Lake Whitefish) and nylon-monofilament (Lake Whitefish) gill nets, so we corrected effort data for species- and lake-specific changes in gill net material to develop standardized indices of CPE. Corrections relied on the methods of Selgeby (1982) for Cisco in state of Wisconsin waters of Lake Superior. We based temporal gear conversions on efficiency curves developed from the best available data on annual proportion of each gill net material used to target Cisco and Lake Whitefish in each lake and the relative efficiency of each material (see Figure S1, Supplemental Material; Hile and Buettner 1955; Pycha 1962; Cucin and Regier 1965; Berst and Spangler 1973; Wells and McLain 1973; Selgeby 1982). We applied efficiency curves to all effort data across all lakes in each year to provide standardized indices of effort and thus CPE. Nylon-multifilament gill nets were shown to be 2.5-fold more efficient for Cisco and 3.0-fold more efficient for Lake Whitefish than linen and cotton gill nets (Lawler 1950; McCombie and Fry 1960; Selgeby 1982). Similarly, nylon-monofilament gill nets were shown to be 1.8-fold more efficient for Lake Whitefish than nylon-multifilament gill nets (Collins 1979). We expressed CPE as linen and cotton gill net equivalents (Figure 3).

Despite the previously described deficiencies associated with fishery-dependent data (see Harley et al. 2001; Walters 2003), we assumed CPE to be a reasonable approximation of underlying abundance for both Cisco and Lake Whitefish because the mesh sizes used for each species were generally consistent during 1929–1970 and targeted most of the age-classes present in populations at that time (Van Oosten 1929, 1939; Smith 1956; Roelofs 1958; Dryer and Beil 1964; Cucin and Regier 1965; Jensen 1976). We removed CPE estimates corresponding to annual totals of effort outside the 2.5–97.5 percentile range, based on all available unit- and species-specific effort data pooled (ca. 10% of all CPE estimates), to ensure that individual data points used for analyses (i.e., paired Cisco–Lake Whitefish CPE) were not based on unusually high or low levels of targeted effort for either species. Unit-specific annual totals of effort used for analyses were log-normally distributed and ranged 5–15,174 km of net (median = 266 km of net) for Cisco and 6–15,126 km of net (median = 946 km of net) for Lake Whitefish (Jensen and Buettner 1976), which was far greater than any possible fishery-independent sampling effort. For comparison purposes, unit-specific annual totals of effort from recent (2015–2019) fishery-independent gill net assessments ranged 1–13 km of net (median = 5 km of net) for Lake Whitefish in state of Michigan waters of Lake Michigan (preliminary agency data provided by T. Treska, USFWS; see Data S2, Supplemental Material). All remaining CPE estimates were log_e-transformed to meet assumptions of normality and translated into standardized Z-scores before analyses (Zar 1999). The relative difference between Cisco and Lake Whitefish CPE ranged 6–706-fold (median = 97-fold) within individual commercial fishery management units, whereas the relative difference between Cisco and Lake Whitefish Z-scores, the values used for analyses, ranged 5–419-fold (median = 63-fold), both of which suggested the relative proportions of each species present within each unit varied substantially throughout the historical period.

Positive and negative correlations between Cisco and Lake Whitefish CPE were tested for significance at a variety of spatial scales throughout all three lakes. A priori spatial models included the following: 1) a 60-parameter global model (GLOBAL) wherein we estimated separate intercepts, correlation coefficients, and variances for each of the 20 commercial fishery management units; 2) a three-parameter reduced model (SIMPLE) wherein we estimated one intercept, correlation coefficient, and variance for all 20 commercial fishery management units combined; 3) a nine-parameter lakewide model (LAKEWIDE) wherein we estimated one intercept, correlation coefficient, and variance for all commercial fishery management units within each lake combined; and 4) 10 different 18–51-parameter regional models (REGIONAL 1–10) wherein we estimated separate intercepts, correlation coefficients, and variances for regional commercial fishery management unit groups within each lake (Table 1). Regional models included a variety of a priori nearest-neighbor and regional commercial fishery management unit groups based on trends in CPE for each species.

The three best-fit spatial models—LAKEWIDE, REGIONAL 10, and SIMPLE—all had similar levels of support (ΔAIC_c < 3.0; see Table 2; Burnham and Anderson 2002), and we used these models to determine whether there was a significant correlation between Cisco and Lake Whitefish CPE (positive and negative). We evaluated pairwise (Cisco–Lake Whitefish) comparisons using the Pearson product-moment correlation coefficient ρ (Zar 1999) and a model-wide P value of 0.05 with a Bonferroni correction (i.e., pairwise P ≤ 0.05/total number of pairwise comparisons; Miller 1966). For example, because only one pairwise comparison would be made for the three-parameter reduced model (SIMPLE), we applied no correction factor to the model-wide P value of 0.05 (P ≤ 0.05 for all commercial fishery management units combined). By contrast, because 20 pairwise comparisons would be made for the 60-parameter global model (GLOBAL), we applied a correction factor of 20 to the model-wide P value of 0.05 (P ≤ 0.05/20 for each commercial fishery management unit). Although significant correlations between Cisco and Lake Whitefish CPE do not imply cause and effect (Zar 1999), we would expect negative correlations if competition or predation were limiting one species, whereas we would expect positive correlations if the same large-scale environmental variable, such as temperature, had the same effect (positive or negative) on both species. Relative proportions of positive and negative correlations were provided for the three best-fit models for all pairwise comparisons (0.0 ≤ |ρ| ≤ 1.0) and those with strong correlations (0.5 ≤ |ρ| ≤ 1.0; Zischke et al. 2017).

The three best-fit spatial models included the following: 1) a nine-parameter lakewide model (LAKEWIDE; ω_i = 0.59) wherein we estimated one intercept, correlation coefficient, and variance for all commercial fishery management units within each lake combined; 2) a 27-parameter regional model (REGIONAL 10; ω_i = 0.23) wherein we estimated separate intercepts, correlation coefficients, and variances for three east–west (Superior) or north–south (Michigan and Huron) regional commercial fishery management unit groups within each lake; and 3) a three-parameter reduced model (SIMPLE; ω_i = 0.14) wherein we estimated one intercept, correlation coefficient, and variance for all 20 commercial fishery management units combined (Figure 1; Tables 1 and 2). There was either no correlation between Cisco and Lake Whitefish CPE or a positive correlation for most (12 of 13) pairwise comparisons (Figures 4–6). We identified no strong positive or negative correlations in the lakewide (LAKEWIDE) or reduced (SIMPLE) models (Figures 4 and 6). In the regional model (REGIONAL 10), we identified strong and positive correlations between Cisco and Lake Whitefish CPE in two regions (ρ = 0.71, n = 41, P < 0.006 [regional group = MS-1]; ρ = 0.59, n = 78, P < 0.006 [regional group = MM-7 and MM-8]) and a weak negative correlation in one region (ρ = −0.45, n = 48, P < 0.006 [regional group = MH-2, MH-3, and MH-4]; Figure 5). The three best-fit models—LAKEWIDE, REGIONAL 10, and SIMPLE—had a combined ω_i of 0.97, whereas that for all other models (GLOBAL and REGIONAL 1–9) was only 0.03 (Table 2).

Collectively, our findings suggest that Cisco and Lake Whitefish CPE were largely independent of each other (see Figures 4–6); thus, these species likely did not interact to the detriment of one another in Michigan waters of the upper Great Lakes during 1929–1970. Our findings were consistent with a similar study for South Bay, Lake Huron, where most (three of four) pairwise correlations suggested no interaction between Cisco and Lake Whitefish during 1949–1984 (Henderson and Fry 1987). Similarly, with the use of multivariate time series models to examine commercial gill net catches in province of Ontario waters of Lake Superior during 1948–1958 and 1963–1973, best-fit models for each period suggested that intraspecific interactions were more important than interspecific interactions between these species (Stone and Cohen 1990). Our findings also were consistent with recent findings that suggest environmental variability, and to a lesser extent compensatory density dependence, drives Cisco and Lake Whitefish abundances throughout the upper Great Lakes (e.g., Rook et al. 2012, 2013; Lynch et al. 2015; Myers et al. 2015), both of which would minimize the effects of interspecific interactions, such as competition or predation (e.g., Todd and Davis 1995; Carl and McGuiness 2006; Stockwell et al. 2014). Heavy exploitation targeting both species (Baldwin et al. 2018) could have had a similar effect by maintaining abundances below levels required for significant interspecific interactions (e.g., Skud 1982). As a result, historical evidence of interactions between Cisco and Lake Whitefish may be limited, both in our study and previous studies.

Our findings were correlative, and as previously stated, cannot establish cause and effect (Zar 1999). However, visual inspection of CPE for each species plotted over time can provide insight into potential mechanisms driving correlations identified in our study (see Figures 4–6). For example, strong positive correlations (western Lake Superior [regional group = MS-1] and southern Lake Michigan [regional group = MM-7 and MM-8]) were associated with concurrent declines in CPE followed by no recovery for both species, whereas the only negative correlation (central Lake Huron [regional group = MH-2, MH-3, and MH-4]) was associated with concurrent declines in CPE for both species followed by no recovery for Cisco and full recovery for Lake Whitefish (see Figure 1; Figure S2, Supplemental Material). Previous studies suggest that overharvest and interactions with exotic species likely drove initial declines in CPE for Cisco and Lake Whitefish throughout the upper Great Lakes, whereas eutrophication in Saginaw Bay also played a key role in Lake Huron (Selgeby 1982; Fleischer 1992; Madenjian et al. 2008, 2011; Gorman 2012). Given that nearly 70% of Cisco and only about 10% of Lake Whitefish harvested in state of Michigan waters of Lake Huron during 1929–1970 were from Saginaw Bay (Baldwin et al. 2018), the one negative correlation identified in our study was likely driven by limited Cisco reproduction (e.g., Madenjian et al. 2008, 2011), concurrent with the early stages of Lake Whitefish recovery, rather than competition or predation between these species. However, because our findings were correlative, we cannot exclude the possibility of negative interactions between Cisco and Lake Whitefish on a regional scale.

Our findings are subject to three main caveats. First, our analyses did not account for potential effects of nonstationarity (Haddon 2011) on correlations between Cisco and Lake Whitefish CPE. As a result, our findings are “snapshots in time” that represent average conditions during 1929–1970. Visual inspection of lakewide trends in CPE for both species suggested the possibility of a break-point in all three lakes during the mid-1950s (see Figure 3). Although these break-points were likely related to extirpation or near extirpation of Cisco (see Eshenroder et al. 2016), we cannot exclude the possibility of a “regime shift” that fundamentally changed coregonine population dynamics in each lake, as observed for the closely related Bloater Coregonus hoyi in Lake Michigan (Bunnell et al. 2006). Second, our findings are also “snapshots in space” that represent conditions in Michigan waters of the upper Great Lakes. Commercial fishery management units used in our study are human constructs based largely on political jurisdictions (Smith et al. 1961), and both Cisco and Lake Whitefish are known to move among these units (e.g., Smith and Van Oosten 1940). Although our findings likely apply to most other areas of the upper Great Lakes, we encourage readers to exercise caution when interpreting our results, because they may not apply to all jurisdictions and we did not consider movement among commercial fishery management units for either species. Third, because we used commercial gill net CPE for all analyses, our findings are subject to all the normal caveats associated with fishery-dependent data. Despite making every attempt to account for spatial differences and known changes in catchability (i.e., changes in gill net material), factors such as species bias, fishing patterns, and weather could have influenced our findings (Liu and Jensen 1992). Therefore, we encourage readers to exercise caution when interpreting our results.

In general, previous studies of interactions between Cisco and Lake Whitefish in the upper Great Lakes were mesocosm experiments (e.g., Todd and Davis 1995) or field studies with limited spatial and temporal scales (e.g., Stockwell et al. 2014; Eckert et al. 2018). Therefore, there is a need for future large-scale studies to evaluate potential interactions between these species under current environmental and ecological conditions in each lake, conditions that have changed substantially since 1970 (e.g., Nalepa et al. 2005; Ebener et al. 2008b; Barbiero et al. 2012; Dove and Chapra 2015; Ives et al. 2019). Future studies focusing on previously identified or newly hypothesized mechanisms that could lead to negative interactions between Cisco and Lake Whitefish, (such as competition during the larval stage [e.g., Todd and Davis 1995], Lake Whitefish predation on Cisco eggs [e.g., Stockwell et al. 2014], Cisco predation on Lake Whitefish larvae [e.g., Carl and McGuiness 2006], and differential foraging abilities or predation from other species because of differences in size during the larval stage [e.g., Eckert et al. 2018]), would be useful for coregonine management, but should be scaled to accommodate environmental stochasticity (interannual variation) and species separation (segregation in time or space). Long-term monitoring of large-scale “natural experiments,” combined with an ecosystem-based modeling approach (e.g., Kitchell et al. 2000), may provide opportunities to evaluate potential interactions between Cisco and Lake Whitefish that cannot or should not be evaluated in small-scale laboratory or field studies (Diamond 1983). The USFWS, at the request of the Lake Huron Committee of the Great Lakes Fishery Commission, has begun to stock at least 750,000 fall fingerling Cisco per year into Saginaw Bay until 2027 (R. Gordon, USFWS, personal communication), which may provide opportunities for further investigation. Given that strong Cisco year-classes can persist in Lake Superior for more than 20 y (Ebener et al. 2008a; Stockwell et al. 2009), long-term monitoring of this “natural experiment” may be possible for the next several decades.

Please note: The Journal of Fish and Wildlife Management is not responsible for the content or functionality of any supplemental material. Queries should be directed to the corresponding author.

Data S1. Data for efficiency curves and analyses used to determine whether there was a correlation between Cisco Coregonus artedi and Lake Whitefish Coregonus clupeaformis commercial gill net catch per effort (CPE; kilograms per kilometer of net, positive and negative) in state of Michigan waters of lakes Superior, Michigan, and Huron during 1929–1970. This file contains columns for 1) lake (LAKE); 2) commercial fishery management unit (MU); 3) year (YEAR); 4) gill net material used to target Lake Whitefish (WF_MAT; LINCOT = linen and cotton, LINCOT_MULT = mix of linen and cotton and nylon-multifilament, MULT = nylon-multifilament, and MULT_MONO = mix of nylon-multifilament and nylon-monofilament); 5) gill net material used to target Cisco (CS_MAT; LINCOT = linen and cotton, LINCOT_MULT = mix of linen and cotton and nylon-multifilament, and MULT = nylon-multifilament); 6) gill net conversion factors (multipliers) used to convert effort into linen and cotton gill net equivalents for Lake Whitefish (WF_CF); 7) gill net conversion factors (multipliers) used to convert effort into linen and cotton gill net equivalents for Cisco (CS_CF); 8) catch (kilograms) of Lake Whitefish (WF_KG); 9) catch (kilograms) of Cisco (CS_KG); 10) corrected effort (kilometers of net) for Lake Whitefish (WF_KM); 11) corrected effort (kilometers of net) for Cisco (CS_KM); 12) corrected CPE for Lake Whitefish (WF_CPE); 13) corrected CPE for Cisco (CS_CPE); 14) log_e-transformed corrected CPE for Lake Whitefish (LN_WF_CPE); 15) log_e-transformed corrected CPE for Cisco (LN_CS_CPE); 16) standardized Z-scores (the values used for analyses) based on log_e-transformed corrected CPE for Lake Whitefish (WF_Z); 17) standardized Z-scores (the values used for analyses) based on log_e-transformed corrected CPE for Cisco (CS_Z); and 18) whether individual data points were removed (REMOVE) or kept (KEEP) before analyses based on values of corrected effort for either species (i.e., WF_KM or CS_KM) that fell outside the 2.5–97.5 percentile range for all available unit- and species-specific effort data pooled (E_FILTER). We expressed all corrected data as linen and cotton gill net equivalents. To express data as linen and cotton gill net equivalents, we multiplied nylon-monofilament and nylon-multifilament gill net effort by species-, lake-, and year-specific gill net conversion factors (WF_CF and CS_CF). Generation of efficiency curves for each species and lake is possible by plotting gill net conversion factors over time (i.e., WF_CF and CS_CF vs. YEAR; see Figure S1). It is worth noting that gill net conversion factors (WF_CF and CS_CF) were based on data collected before increases in water clarity throughout all three lakes (Dove and Chapra 2015) and relative efficiencies may have changed in recent years (Fera et al. 2015). Although this does not affect our findings, we encourage readers to exercise caution when applying these conversion factors to more recent data. The data are equivalent to data provided in the official U.S. Geological Survey data release (Rook 2021).

Available: https://doi.org/10.3996/JFWM-20-062.S1 (140 KB XLSX)

Data S2. Annual totals of effort (kilometers of net) from recent (2015–2019) fishery-independent gill net assessments targeting Lake Whitefish Coregonus clupeaformis within individual commercial fishery management units in state of Michigan waters of Lake Michigan (data provided by T. Treska, U.S. Fish and Wildlife Service). This file contains columns for 1) lake (LAKE); 2) commercial fishery management unit (MU); 3) year (YEAR); and 4) effort (EFFORT_KM). Multiple state, federal, and tribal agencies collected data; they represent nylon-monofilament gill net equivalents and are intended for comparison purposes only. All values are preliminary and have not gone through the official U.S. Geological Survey data release process. Interested readers can contact the U.S. Fish and Wildlife Service for further information. See Figure 1 for locations of commercial fishery management units in state of Michigan waters of Lake Michigan and Smith et al. (1961) for detailed descriptions of all commercial fishery management units in the Great Lakes.

Available: https://doi.org/10.3996/JFWM-20-062.S2 (1 KB CSV)

Figure S1. Data for efficiency curves used to determine whether there was a correlation between Cisco Coregonus artedi and Lake Whitefish Coregonus clupeaformis commercial gill net catch per effort (CPE; kg/km of net, positive and negative) in state of Michigan waters of lakes Superior, Michigan, and Huron during 1929–1970. Curves represent estimated relative efficiency compared with the previously used gill net material (i.e., nylon-multifilament compared with linen and cotton or nylon-monofilament compared with nylon-multifilament) based on the best available data on annual proportion of each gill net material used to target Cisco and Lake Whitefish in each lake and the relative efficiency of each material (Hile and Buettner 1955; Pycha 1962; Cucin and Regier 1965; Berst and Spangler 1973; Wells and McLain 1973; Selgeby 1982). For Cisco, the transition from linen and cotton to nylon-multifilament gill nets occurred during 1945–1961. For Lake Whitefish, the transition from linen and cotton to nylon-multifilament gill nets occurred during 1945–1953, whereas the transition from nylon-multifilament to nylon-monofilament gill nets (Multi to Mono) occurred during 1965–1976. We generated efficiency curves for each species and lake by plotting gill net conversion factors over time (i.e., WF_CF and CS_CF vs. YEAR; see Data S1). It is worth noting that gill net conversion factors (WF_CF and CS_CF; see Data S1) were based on data collected before increases in water clarity throughout all three lakes (Dove and Chapra 2015) and relative efficiencies may have changed in recent years (Fera et al. 2015). Although this does not affect our findings, we encourage readers to exercise caution when applying these conversion factors to more recent data.

Available: https://doi.org/10.3996/JFWM-20-062.S3 (79 KB DOCX)

Figure S2. Cisco Coregonus artedi and Lake Whitefish Coregonus clupeaformis commercial gill net catch per effort (CPE; kilograms per kilometer of net) for selected commercial fishery management units in state of Michigan waters of lakes Superior, Michigan, and Huron during 1929–1970 (Jensen and Buettner 1976). Commercial fishery management units are Superior (MS), Michigan (MM), and Huron (MH). Data are linen and cotton gill net equivalents and represent the two strong positive correlations (western Lake Superior [regional group = MS-1] and southern Lake Michigan [regional group = MM-7 and MM-8]) and the one negative correlation (central Lake Huron [regional group = MH-2, MH-3, and MH-4]) identified in the three best-fit spatial models (LAKEWIDE, REGIONAL 10, and SIMPLE). The y-axes (log_e scale) are identical to show differences in CPE between commercial fishery management units and species. Strong positive correlations were associated with concurrent declines in CPE followed by no recovery for both species, whereas the only negative correlation was associated with concurrent declines in CPE for both species followed by no recovery for Cisco and full recovery for Lake Whitefish. See Table 1 for descriptions of all spatial models. See Table 2 for model rankings according to scaled second-order Akaike Information Criterion (Burnham and Anderson 2002). See Figure 1 for locations of commercial fishery management units and regional commercial fishery management unit groups for the second-ranked model (REGIONAL 10). See Figures 4–6 for all correlation results for the three best-fit models (LAKEWIDE, REGIONAL 10, and SIMPLE).

Available: https://doi.org/10.3996/JFWM-20-062.S4 (414 KB DOCX)

Reference S1. Berst AH, Spangler GR. 1973. Lake Huron: the ecology of the fish community and man's effect on it. Ann Arbor, Michigan: Great Lakes Fishery Commission, Technical Report Number 21.

Available: https://doi.org/10.3996/JFWM-20-062.S5 (957 KB PDF) and http://www.glfc.org/pubs/TechReports/Tr21.pdf

Reference S2. Brege DA, Kevern NR. 1978. Michigan commercial fishing regulations: a summary of Public Acts and Conservation Commission orders, 1865 through 1975. Ann Arbor, Michigan: Michigan Sea Grant Program, Reference Report MICHU-SG-78-605.

Available: https://doi.org/10.3996/JFWM-20-062.S6 (2.37 MB PDF)

Reference S3. Bronte CR, Bunnell DB, David SR, Gordon R, Gorsky D, Millard MJ, Read J, Stein RA, Vaccaro L. 2017. Report from the workshop on coregonine restoration science. Reston, Virginia: United States Geological Survey, Open-File Report 2017–1081.

Available: https://doi.org/10.3996/JFWM-20-062.S7 (924 KB PDF) and https://pubs.usgs.gov/of/2017/1081/ofr20171081.pdf

Reference S4. Ebener MP. 2013. Status of whitefish and ciscoes. Pages 29–35 in Riley SC, editor. The state of Lake Huron in 2010. Ann Arbor, Michigan: Great Lakes Fishery Commission, Special Publication 13-01.

Available: https://doi.org/10.3996/JFWM-20-062.S8 (1.56 MB PDF) and http://www.glfc.org/pubs/SpecialPubs/Sp13_01.pdf

Reference S5. Ebener MP, Stockwell JD, Yule DL, Gorman OT, Hrabik TR, Kinnunen RE, Mattes WP, Oyadomari JK, Schreiner DR, Geving S, Scribner K, Schram ST, Seider MJ, Sitar SP. 2008a. Status of Cisco (Coregonus artedi) in Lake Superior during 1970–2006 and management and research considerations. Ann Arbor, Michigan: Great Lakes Fishery Commission, Lake Superior Technical Report Number 1.

Available: https://doi.org/10.3996/JFWM-20-062.S9 (1.18 MB PDF) and http://www.glfc.org/pubs/lake_committees/superior/Cisco.pdf

Reference S6. Eshenroder RL, Vecsei P, Gorman OT, Yule DL, Pratt TC, Mandrak NE, Bunnell DB, Muir AM. 2016. Ciscoes (Coregonus, subgenus Leucichthys) of the Laurentian Great Lakes and Lake Nipigon. Ann Arbor, Michigan: Great Lakes Fishery Commission.

Available: https://doi.org/10.3996/JFWM-20-062.S10 (33.31 MB PDF) and www.glfc.org/pubs/misc/Ciscoes_of_the_Laurentian_Great_Lakes_and_Lake_Nipigon.pdf

Reference S7. Hile R. 1962. Collection and analysis of commercial fishery statistics in the Great Lakes. Ann Arbor, Michigan: Great Lakes Fishery Commission, Technical Report Number 5.

Available: https://doi.org/10.3996/JFWM-20-062.S11 (531 KB PDF) and http://www.glfc.org/pubs/TechReports/Tr05.pdf

Reference S8. Hile R, Buettner HJ. 1955. Commercial fishery for chubs (ciscoes) in Lake Michigan through 1953. Washington, D.C.: U.S. Fish and Wildlife Service, Special Scientific Report–Fisheries Number 163.

Available: https://doi.org/10.3996/JFWM-20-062.S12 (2.38 MB PDF) and https://spo.nmfs.noaa.gov/sites/default/files/legacy-pdfs/SSRF163.pdf

Reference S9. Jensen AL, Buettner HJ. 1976. Lake Trout, whitefish, chubs, and Lake Herring yield and effort data for State of Michigan waters of the upper Great Lakes: 1929–1973. Ann Arbor, Michigan: University of Michigan Sea Grant, Technical Report Number 52.

Available: https://doi.org/10.3996/JFWM-20-062.S13 (4.67 MB PDF)

Reference S10.Lake Huron Technical Committee (LHTC). 2007. Strategy and options for promoting the rehabilitation of Cisco in Lake Huron. Ann Arbor, Michigan: Great Lakes Fishery Commission.

Available: https://doi.org/10.3996/JFWM-20-062.S14 (146 KB PDF) and http://www.glfc.org/pubs/lake_committees/huron/LakeHuron_CiscoRehab.pdf

Reference S11. Nalepa TF, Mohr LC, Henderson BA, Madenjian CP, Schneeberger PJ. 2005. Lake Whitefish and Diporeia spp. in the Great Lakes: an overview. Pages 3–19 in Mohr LC, Nalepa TF, editors. Proceedings of a workshop on the dynamics of Lake Whitefish (Coregonus clupeaformis) and the amphipod Diporeia spp. in the Great Lakes. Ann Arbor, Michigan: Great Lakes Fishery Commission, Lake Superior Technical Report 66.

Available: https://doi.org/10.3996/JFWM-20-062.S15 (3.84 MB PDF) and http://www.glfc.org/pubs/TechReports/Tr66.pdf

Reference S12. Smith SH, Buettner HJ, Hile R. 1961. Fishery statistical districts of the Great Lakes. Ann Arbor, Michigan: Great Lakes Fishery Commission, Technical Report Number 2.

Available: https://doi.org/10.3996/JFWM-20-062.S16 (453 KB PDF) and http://www.glfc.org/pubs/TechReports/Tr02.pdf

Reference S13. Wells L, McLain AL. 1973. Lake Michigan: man's effects on native fish stocks and other biota. Ann Arbor, Michigan: Great Lakes Fishery Commission, Technical Report Number 20.

Available: https://doi.org/10.3996/JFWM-20-062.S17 (838 KB PDF) and http://www.glfc.org/pubs/TechReports/Tr20.pdf

This project was funded by the Great Lakes Restoration Initiative, U.S. Environmental Protection Agency under the DOI Coregonine Restoration Template. The USFWS (old Bureau of Commercial Fisheries) made this project possible by collecting historical commercial fishery data. We are grateful to A. Jensen and H. Buettner for compiling catch and effort data used for this project. We thank M. Slattery for geographic information system and mapping work. We also thank the Associate Editor and anonymous reviewers for substantially improving the manuscript.

Any use of trade, product, website, or firm names in this publication is for descriptive purposes only and does not imply endorsement by the U.S. Government.